Homogenization of a boundary value problem for the \(n\)-Laplace operator on a \(n\)-dimensional domain with rapidly alternating boundary condition type: the critical case (Q2313381)

The authors describe the asymptotic behavior of the solution \(u_{\varepsilon }\) to the \(n\)-Laplace problem: \(-\mathrm{div} (\left\vert \nabla u_{\varepsilon }\right\vert ^{n-2}\nabla u_{\varepsilon })=f\) posed in a bounded domain \( \Omega \subset \mathbb{R}^{n}\cap \{x_{1}>0\}\), \(n\geq 3\), with the boundary conditions \(\left\vert \nabla u_{\varepsilon }\right\vert ^{n-2}(\nabla u_{\varepsilon },\nu )+\beta (\varepsilon )\sigma (x,u_{\varepsilon })\) on \(\gamma _{\varepsilon }\), \(\left\vert \nabla u_{\varepsilon }\right\vert ^{n-2}(\nabla u_{\varepsilon },\nu )=0\) on \( \Gamma _{1}\setminus \overline{\gamma }_{\varepsilon }\), \(u_{\varepsilon }=0\) on \(\Gamma _{2}\), the boundary \(\partial \Omega \) of \(\Omega \) being decomposed as \(\Gamma _{1}\cup \Gamma _{2}\) where \(\Gamma _{1}=\partial \Omega \cap \{x_{1}=0\}\), \(\gamma _{\varepsilon }\) being the union of \( \varepsilon \)-periodically distributed intervals of length \(a_{\varepsilon }\). Here \(f\in L_{q}(\Omega )\), \(q=\frac{n}{n-1}\), and \(\sigma \) is a continuously differentiable function on \(\overline{\Omega }\times \mathbb{R}\) such that \(\sigma (x,0)=0\), \((\sigma (x,u)-\sigma (x,v))(u,v)\geq k_{1}(u-v)^{2}\) and \(\left\vert \sigma (x,u)\right\vert \leq k_{2}\left\vert u\right\vert ^{n-1}\) for every \(u,v\in \mathbb{R}\). The parameters \( a_{\varepsilon }\) and \(\beta (\varepsilon )\) satisfy \((\beta (\varepsilon ))^{1/(n-1)}a_{\varepsilon }\varepsilon ^{-1}\rightarrow _{\varepsilon \rightarrow 0}C_{0}^{2}\) and \[ \frac{1}{(\beta (\varepsilon ))^{1/(n-1)}a_{\varepsilon }\ln (4a_{\varepsilon }\varepsilon ^{-1})} \rightarrow _{\varepsilon \rightarrow 0}C_{1}^{2}. \] The authors first build the variational formulation associated to this nonlinear problem. They prove the existence of a unique weak solution which satisfies the uniform estimates \(\left\Vert \nabla u_{\varepsilon }\right\Vert _{L_{n}(\Omega )}\leq K\) and \(\beta (\varepsilon )\left\Vert u_{\varepsilon }\right\Vert _{L_{n}(\gamma _{\varepsilon })}^{n}\leq K\). The main result proves that this weak solution weakly converges in \(W^{1,n}(\Omega ,\Gamma _{2})\) and strongly in \(L_{n}(\Omega )\) to the solution \(u_{0}\) of the problem \(- \mathrm{div}(\left\vert \nabla u_{0}\right\vert ^{n-2}\nabla u_{0})=f\) in \(\Omega \), with the boundary conditions \(\left\vert \nabla u_{0}\right\vert ^{n-2}\partial _{x_{1}}u_{0}=\mathcal{A}\left\vert H(x,u_{0})\right\vert ^{n-2}H(x,u_{0})\) on \(\Gamma _{1}\), \(u_{0}=0\) on \(\Gamma _{2}\), where \( \mathcal{A}=2^{-1}C_{0}^{2(n-1)}C_{1}^{2(n-1)}\) and \(H\) is defined through \( \mathcal{B}\left\vert H\right\vert ^{n-2}H=\sigma (x,u-H)\), with \(\mathcal{B} =C_{1}^{2(n-1)}\omega _{n}/(2\left\vert \gamma _{0}\right\vert )\). The proof is essentially based on the construction of local test-functions \( W_{\varepsilon }\) and \(Q_{\varepsilon }\) and on their properties which allow to pass to the limit in the variational formulation associated to the nonlinear problem.